How to read a microphone test: levels, peaks and background noise explained

When you speak into a microphone, the browser or operating system turns the sound into a stream of numbers. A test page reads those numbers many times per second and draws them as a moving bar. That bar is not 'volume' in the way a speaker has a volume knob. It is the strength of the signal that reached your computer, after the microphone, the preamplifier and any gain setting have all done their work.

That is why the same voice can look very different on two computers, or on the same computer after you change one setting. A test is a snapshot of the whole chain, not just of the microphone.

Digital audio has a hard ceiling called full scale, written 0 dBFS. Nothing can be louder than that, so every other level is a negative number: -6 dBFS is a little below the ceiling, -40 dBFS is very quiet. If a test page shows a bar instead of numbers, the right edge of the bar is that ceiling.

The decibel scale is logarithmic, so small numeric steps matter. Going from -12 dBFS to -6 dBFS is a doubling of the signal's amplitude, which is why a bar that looks only a little longer can sound noticeably louder.

Speech is made of loud bursts (the 'p', 't' and 'k' sounds, a laugh, a raised voice) sitting on top of a much lower average. A good meter shows both. The average tells you how loud you sound to a listener. The peak tells you whether any single moment came close to the ceiling.

A useful rule of thumb for spoken voice is to keep peaks somewhere between about -12 and -6 dBFS when you talk at your normal volume, with the average sitting well below that. That leaves headroom for the moments you get louder without warning, and it is high enough that the signal is far above the electronic hiss every microphone and preamp produces.

If the bar hits the very end, or a 'clip' or 'peak' indicator lights up, the signal has been cut off flat at the ceiling. This is clipping, and it cannot be repaired afterwards, because the information that was above the ceiling was never recorded. It sounds like crackling or harshness on loud syllables.

The fix is almost always to lower the gain (see below), not to speak more quietly. If you need to be loud, for example when streaming a game, set your levels while being loud, not while whispering.

Stop talking for a few seconds and look at the meter. In a perfectly silent system the bar would sit at the left edge. In reality there is always something: fan noise, air conditioning, traffic, the hiss of the preamplifier, electrical interference from a charger or a monitor.

That resting level is the noise floor, and it matters more than the loudest peak, because it is always present underneath your voice. The gap between your speech and your noise floor is the signal-to-noise ratio. A voice that peaks around -9 dBFS over a noise floor near -60 dBFS has a wide, comfortable gap. If the floor sits at -35 dBFS, the room or the gain setting is the problem, not the microphone.

Most systems give you more than one control for how loud the microphone is, and they are not interchangeable. The input volume slider in your operating system's sound settings sets how strongly the signal is amplified. Some Windows drivers also expose a separate 'microphone boost' in decibel steps, which adds amplification at the front of the chain, together with the microphone's own noise.

Boost is useful for a genuinely quiet microphone, but it raises the hiss along with the voice. Use the smallest amount that gets your peaks into the target range. Many USB microphones have a physical gain knob, and that is usually the best place to adjust, because it works before the signal is converted to digital.

Browsers and call apps often apply automatic gain control, echo cancellation and noise suppression to the microphone before you ever see a level. These features can make a test look better than your raw setup really is: the bar jumps to a healthy level even though the microphone is far away, and quiet background noise disappears the moment you stop talking.

That is helpful in a meeting and misleading when you are trying to diagnose a problem. If you can, test with the processing features switched off, and again with them on, to see what you get in each situation.

1. Sit where you will normally sit and place the microphone where it will normally be. Distance changes level more than any slider does.

2. Stay quiet for five seconds and note the noise floor. If it is high, deal with the cause first: move away from the fan, close the window, unplug the noisy charger.

3. Speak a full sentence at your normal volume. Watch the peaks, not just the average.

4. Adjust gain so the peaks land around -12 to -6 dBFS and nothing touches the ceiling.

5. Say something loud on purpose, like a laugh or an emphatic word. If it clips, lower the gain a little and repeat.

The test meter shows what reaches your computer. It cannot show what the other person hears. If listeners still say you sound distant, echoey or robotic, the cause is usually outside the level: a room with hard walls and no soft furniture, a laptop microphone far from your mouth, Bluetooth headphones that switch to a lower quality mode when the microphone is on, or the call app's own compression.

In that case, record a short clip, play it back with headphones and listen for room echo and background noise. What you hear in that recording is much closer to what your listeners get than any number on a meter.